A text mining device has been known from before that can find multiple texts representing the same characteristic content despite having mutually different expressions, from among multiple input texts (for example, see Patent Literature 1). This text mining device mutually associates and stores multiple differing expressions having the same characteristic content. This text mining device is such that when an expression associated with a prescribed expression is included in input text, the expression is converted to the prescribed expression and text containing the prescribed expression is found.